VidHOI

VideosIntroduced 2021-05-25

VidHOI is a video-based human-object interaction detection benchmark. VidHOI is based on VidOR which is densely annotated with all humans and predefined objects showing up in each frame. VidOR is also more challenging as the videos are non-volunteering user-generated and thus jittery at times.

Image source: https://xdshang.github.io/docs/vidor.html

Benchmarks

Human-Object Interaction Anticipation/Person-wise Top5: t=1(mAP@0.5)Human-Object Interaction Anticipation/Person-wise Top5: t=3(mAP@0.5)Human-Object Interaction Anticipation/Person-wise Top5: t=5(mAP@0.5)Human-Object Interaction Detection/Detection: Full (mAP@0.5)Human-Object Interaction Detection/Detection: Non-Rare (mAP@0.5)Human-Object Interaction Detection/Detection: Rare (mAP@0.5)Human-Object Interaction Detection/Oracle: Full (mAP@0.5)Human-Object Interaction Detection/Oracle: Non-Rare (mAP@0.5)Human-Object Interaction Detection/Oracle: Rare (mAP@0.5)